RESEARCH Open Access Effects of cattle rustling and household characteristics on migration decisions and herd size amongst pastoralists in Baringo District, Kenya George K Kaimba 1 , Bernard K Njehia 2,4 and Abdi Y Guliye 3* * Correspondence: guliye@egerton. ac.ke 3 Department of Animal Sciences, Egerton University, P. O. Box 536- 20115, Egerton, Kenya Full list of author information is available at the end of the article Abstract Pastoral communities in arid and semi-arid lands (ASALs) of Kenya depend on livestock for their livelihood. However, these ASALs are charact erized by temporal and spatial climatic variation, making availability of resources uneven. Mobility is a key strategy used by pastoralists to efficiently utilize available resources, notably pasture and water. This strategy is being interrupted by a vicious cycle of livestock rustling/raiding. This study was conducted to elucidate the effects of livestock rustling and other household characteristics on migration decisions and herd size amongst pastoralists in Baringo District in Kenya. A sample of 110 pastoralists were interviewed using a structure d questionnaire. Binary probit model was used to explain the probability of migrating while ordinary least square was used to explain effects on herd size. Gender and age of the household head are significant (P < 0.1 and P < 0.05, respectively) determinants of migration, whereas both also significantly (P < 0.1) influenced herd size. Intensity of rustling, and loss of livestock to drough t and/or disease also significantly (P < 0.01) influence the decision to migrate. Level of education had significant (P < 0.1) and negative influence on herd size, whereas size of household had significant (P < 0.01) and positive impact on herd size. Non- livestock income had significa nt (P < 0.05) and negative influence on migration and herd size. The practice of livestock rustling, rampant amongst pastoralist communities in Kenya and sometimes occurs across borders, influences pastoralists’ decision to migrate and also their herd sizes. It destabilizes communities and undermines their normal livelihood strategies, thus contributing to increased poverty. Increasing the level of development in pastoral areas and formulation of appropriate policies will help in controlling the rustling menace. Keywords: cattle rustling, migration, herd size, Baringo, Kenya Kaimba et al. Pastoralism: Research, Policy and Practice 2011, 1:18 http://www.pastoralismjournal.com/content/1/1/18 © 2011 K aimba et al; licensee Springer. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Introduction More than 80% of the total l and area in Kenya consists of arid and semi-arid lands (ASALs) (Okoti et al. 2004), where constraining rainfall and temperature conditions pro- vide limited options for sustainable land use, other than mobile livestock rearing. Mobi- lity is the underlying strategy in the utilization of ASALs (Behnke and Scoones 1993), as it enables efficient use of rangeland resources through seasonal migration in search of pasture, water and mineral licks. Thus, seasonal movement and nomadic pastoralism are themajoreconomicactivityandthemainsource of livelihood for the inhabitants of ASAL s. Kenya’s ASALs support more than 30% (approximately 12 million) people, 50% cattle, 70% sheep and goats, and the entire ca mel population (SRA 20 03). It is estimated that the livestock sector provides almost 90% of employment and mo re than 95% of family incomes in Kenya’s ASALs (FAO 2004). Livestock plays multiple roles in the lifestyle of pastoralist s in Kenya, notably as liveli- hood sources, socio-cultural and religious functions, and asset and security against risks (Guliye et al. 2007). For example, livestock is the main source of food by providing milk and meat, the basis of traditional social relations, e.g. payment of dowry (from the groom’s family to the bride’s family) during marriage or compensati on of injured parties in tribal feuds, symbol of prosperity and prestige, store of wealth, and security against drought, disease and other calamities. The pastoralists i n Baringo District of Ken ya are mainly transhumance pastoralists, and they exemplify communities in ASALs that are dependent on livestock for their live- lihood. Traditionally, they move seasonal ly from thei r home bases and drive their herds to places with pasture and water and come back to their homesteads in other seasons when pasture improves. Of all the livestock kept by the Baringo pastoralists, cattle are regarded highly. Because of the importance attached to cattle, the re is a tendency to accumulate them even under unfavourable environmental conditions, often exerting a lot of pressure on the meagre range resources, notably pasture and water. Inevitably, there is competition amongst pastoralists in the district for the available range resources, necessitating frequent livestock movements within the range in search of pasture and water (Raikes 198 1). The occurrence of frequent droughts in ASALs, perhaps a manifes - tation of climate cha nge, contributes to range resource shortages, lea ding to intense competition for the available pasture and wa ter. Thus, mobility remains the key pastoral risk management strategy during times of pasture and w ater shortage. (Little et al. (2001)) point out that pastoralists who migrate with their herds during climatic disasters have consid erably fewer livestock losses than those who do not. However, this mobility in itself causes conflict among the pastoralists due to competition for scarce pasture and water. Pasture and water conflicts have long been part o f the socio-cultural pattern of the pastoral communities in Kenya. The communal land ownership tenure system mostly evident in pa storalist areas provides everyone an equal right of exploi ting the resources. The lands ar e traditional tribal grazing areas, such that migration in search of pasture and water by one tribe into areas that belong to other tribes often causes conflict between pastoralists. Besides, livestock movements into grazing lands and wate ring points that stretch into crop-growing areas al so result in con flicts (Dietz 1987). Over time however, pasture and water around the settled areas steadily decreases, leading to emaciation and loss of livestock. Tra dition ally, whenever scarcity of p asture and water Kaimba et al. Pastoralism: Research, Policy and Practice 2011, 1:18 http://www.pastoralismjournal.com/content/1/1/18 Page 2 of 16 or disease depleted a community’s livestock, it often sought to replenish numbers through raiding/rustling (Mkutu 2000). Livestock rustling/raiding, commonly referred to as cattle rustling in Kenya, involves for- ceful acquisition of livestock (mainly cattle) and is quite common amongst pastoralists in the ASALs of Kenya. Traditionally, cattle rustling often involved small-scale violence and theft of the best livestock or replacement of animals lost through drought or disease. Loss of human lives was rare, and when this occurred, compensation in the form of cattle was paid by the killers’ families to the victims or their families in case of death. However, in recent years, due to proliferation of small arms and commercialization of cattle rustling, there is an emergence of large-scale violent cattle raiding between neighbouring pastoral communities in Kenya (Hendrickson et al. 1996). Moreover, there is an emergence of commercialized cattle rustling where wealthy businessmen, politicians, traders or local people pursuing economic objectives finance raids among the pastoral communities. This greatly interferes with the future and assets of the pastoralists. Consequently, pastoral communities arm themselves for protection against hostile groups. The threats caused by the increasing numbers of human deaths and livestock losses due to cattle rusting and other organised raids probably influences the pastoralists’ mobility and/or their migratory decisions as well as herd size, thereby undermining their asset base and livelihood sources. Thus, besides lack of pasture and water, pastoralist migration could also be influenced by the perceived threats of cattle rusting and the insecurity generated by it (Doss et al. 2008). There is little information on the influence of cattle rustling on migration decisions and herd size of the pastoralists. This study therefore investigated the effects of cattle rustling and other household characteristics on decisions to migrate and herd size amongst pastor- alists in Baringo District, Kenya. Materials and methods Study area The study was conducted in Baringo, one of the arid and semi-arid districts in the Rift Valley Province of Kenya. It is located betwe en latitudes 35°30’ and 36°30’ East and lati- tudes 00°10’ South and 00°140’ North, and covers an area of 10,949 km 2 , of which about 165 km 2 is surface water. The district is hot and dry throughout most of the year. Rain- fall is highly variable, with an annual mean of 635 mm, with weak bimodal peaks recorded from March to May and June to August. The average minimum and maximum temperatures are 20°C and 35°C, respectively. The district is characterised by bare ground and loose sandy loam soil with occasional stones on the surface. Much of the vegetation in the area is Acacia woodland dominated by Acacia tortilis, Acacia reficiens and Boscia corriacea. Other major plant species include Balanites aegyptiaca, Maerua angolensis, Cordia sinensis and Salvadora persica. The district is inhabited by the Pokot, Tugen and Njemps communities whose major occupation is livestock keeping. Sampling procedure The sample population consisted of herders within Baringo District. Data were obtained using multi-stage sampling method. Purposive sampling was used to select the rustling/raiding prone divisions in the district which include Tangulbei, Nginy ang, Marigat, Kallowa and Bartabwa. The selected divisions were used as study clusters (first-stage cluster sampling). Thereafter, locations, within the cluster divisions, were Kaimba et al. Pastoralism: Research, Policy and Practice 2011, 1:18 http://www.pastoralismjournal.com/content/1/1/18 Page 3 of 16 selected at random (second-stage cluster sampling). Then, random samples within each location were selected (third-stage clust er sampling), from which interviews were con- ducted by use of a structured questionnaire. Herders were asked questions about their household characteristics, herd composition, and the level and effects of cattle rustling in the last 5 years. A total of 110 households were selected for interviews from the sampling frame. Secondary data relevant to the study were also obtained and used in the analysis. Theoretical framework This study is based on the theory of risk and uncertainty. It utilizes the possibilities offered by the Dempster-Shafer theory of evidence as one way of representing impre- cise probabilities and partial information in an involuntary decision-making context (Ducey 2001). Pastoral risk management i nvolves making choices/decisions in the face of uncertaintie s. Most of such choices/decisions, including migratory decisions, involve ever yday directly perceptible risks. Such risks are managed instin ctively and int uitively (Adams 1999). Risk is restricted to situations where probabilities are allocated to the occurrence of an event. On the other hand, uncertainty arises when the chances governing stochastic factors are imperfectly known. In this case, a herder contemplat- ing a decision at the height o f cattle rustling would be likely to face both r isk and uncertainty. Just like in many other forms of risks, there is no formal probabilistic assessment done before making a decision to migrate by a pastoralist herder. However, there are two things that are obvious under such circumstances. First, h erders prefer higher social economic status in the community to lower status. Secondly, under uncertainty all herders face the possibility that they would suffer heavy losses, and each must compare what he has to gain against what he has to lose in what would be essen- tially a random draw. Therefore, decisions made due to risk and uncertainties like the fear of cattle rustling or loss of livestock through drought should be able to contend with chances and degree of belief (Ducey 2001). As (Shafer (1976)) points out, if the chance associated with an event is known, i t would be advantageous to adopt those chances as degree of belief and act accordingly. (Caselton and Luo (1992)) recom- mended the utility of Dempster-Shafer theory in decision analysis under risk and uncertainty, particularly where data are sparse and absent. Empirical model The dec ision of the ith herder to migrate depends on unobservable utility index that is determined by the explanatory variables: y i • = β 0 + k  j=1 β j χ ij + u i (1) From Equation 1, the index function can further be expressed as: y i ∗ = β 0 + β 1 x 1i + β 2 x 2i + + β k x ki + u y i ∗ is unobservable but y i =  0ify i ∗<0 1ify i ∗≥0 (2) where y • is a latent variable which is not observed and only the outcome y i (defined as below) is observed. b 0 is a constant and b j are vectors of coefficient to be estimated. Kaimba et al. Pastoralism: Research, Policy and Practice 2011, 1:18 http://www.pastoralismjournal.com/content/1/1/18 Page 4 of 16 The c ki are the i ndependent variables influencing herder i, k are attributes influencing herder i and u i is the error term. Model specification A herder contemplating whether to migrate would have to evaluate whether the ven- ture is worth undertaking or not. The herder’s choice would be based on a set of para- meters or attributes (not necessarily in monetary terms) which describe the suitability of migrating. If X represents a vector of determinants of the decision to migrate, the basic form of the binary probit function with ˇ Z as the predictor variable can be expressed as: ˇ Z = β 0 + β 1 X 1 + β 2 X 2 + +β j X j (3) The decision-making process in this case is unobserved and only the outcome, which is migration, is observable. The prob ability that herder i would choose to migrate can be predicted as: M wl = f  GHH i ,AG i ,ED i , HHS i ,NLI i ,CARUINTY i NCA i ,RSG i ,LO i ,AR i ,BC i P mmd i  (4) where GHH i represents gender of the household head, AG i represents age of the household head, ED i represents education level of the household head measured in terms of number of years in school, HHS i represents the size of the household, CAR- UINTY i is a dummy variable representing cattle rustling intensity in the area, NLI i is the non-livestock income received by herder i, NCA i is the number of cattle o wned by herder i,RSG i is the ratio of sheep and goats to cattle owned by herder i,LO i is a dummy variable representing type of land ownership by herder i,AR i is a dummy vari- able representing whether a herder has lost livestock to cattle rustlers or not, P mmd i is a dummy variable representing herder i’sperceptiononmigrationandBC i is a vector of biophysical characteristics (disease/parasites and drought/famine). To model the impacts of c attle rustling and migration decision on herd size, the study estimated a herd size function using the production function approach as simpli- fied by (Kabubo-Mariara (2003)). Kabubo-Ma riara’s model compared the productivity of private and common property, which is modified in the present study to compare the effect of cattle rustling and migration decisions on herd size, as:  = ν i CRV + β i P i + n  j=1 α ij X j + μ i (5) where F is herd size; CRV represents cattle rustling variables influencing migration decision, (i.e. cattle rustling intensity and whether a herder has lost livestock to cattle rustlers in the past); P i is the predicted probability of migrating from Equation 4; X j is the vector for exogenous variables other than rustling that affect herd size; ν i , a ij and b i are unknown coefficients; and μ i is the stochastic disturbance term. The present study assumed that, other things being constant, decrease in the occur- rence of cattle rustling and positive perceived impact of migration would yield more herd size. Based on this assumption, the herd size model can be specified as: Kaimba et al. Pastoralism: Research, Policy and Practice 2011, 1:18 http://www.pastoralismjournal.com/content/1/1/18 Page 5 of 16 HS i = f ( GHH i AG i ,ED i ,HHS i ,NLI i , CARUINTY i ,LO i ,AR i ,BC i ,INHERIT i ,DOWRY i ,BOUGHT i , PRMIGR i - ) (6) where HS i is the herd size of herder i,INHERIT i is a dummy variable representing whether or not herder i inherited livestock, DOWRY i is a dummy variable representing whetherornotherderi received dowry, BOUGHT i is a dummy variable representing whether or not the herder i bought livestock and PRMIGR i is the predicted probability of migrating estimated in Equation 5. All the other variables are as defined in Equation 4 above. The independent variables used in Equations 4 and 6 of the analysis are sum- marized in Table 1. Statistical analysis STATA software (version 9.0) from StataCorp LP (4905 Lakeway Drive, College Sta- tion, Texas 77845 USA) was used to analyse the data. The estimated Equation 5 above was used in probit analysis of migratory decisions, whereas single-equation ordinary least squares (OLS) estimation was used in the determination of factors influencing herd size. Further, three-stage least square (3SLS) estimation was used to test for simultaneity in t he analysis of the determinants of herd size and the results compared to those of single-equation estimation. T he potential limitations to the analysis that included specification error, omitted variables, simultaneity and heteroscedasticity were taken care of using appropriate econometric procedures. Results The responses of the surveyed pastoralists in Baringo District of Kenya to various household characteristics are presented in Table 2. The results show that most of the households (89%) are headed by males. In the few female -headed households (11%), culture demands that she must consult the oldest son during decision making. Approximately 96% of the p astoralists use family labour as opposed to 4% who use hired labour. The results also indicate that more than 80% of the herders are illiterate. Some of the household heads had at one time enrolled in primary schools but latter dropped out. A few have, however, gone up to secondary school and even beyond. Table 1 Independent variables. Variables Descriptions Units GHH Gender of household head 1 = male, 0 = female AG Age of household head Years ED Education level of household head Years in school HHS Size of the household Number NCA Number of cattle owned Number RSG Ratio of sheep and goats to cattle Number NLI Non-livestock income in the last 5 years KES CARUINTY Intensity of cattle rustling 1 = severe, 2 = moderate LO Type of land ownership 0 = common, 1 = private BC Drought and/or diseases Yes = 1, no = 0 AR Livestock lost to rustlers in the last 5 years Yes = 1, no = 0 INHERIT Livestock inherited Yes = 1, no = 0 DOWRY Dowry received Yes = 1, no = 0 BOUGHT Livestock bought Yes = 1, no = 0 PRMIGR Predicted probability of migrating Number Kaimba et al. Pastoralism: Research, Policy and Practice 2011, 1:18 http://www.pastoralismjournal.com/content/1/1/18 Page 6 of 16 Determinants of pastoral migration decisions The practice of migration by pastoralists with their livestock is an important management strategy used by pastoral communities, aimed at exploiting available range resources. However, in recent times, cattle rustling and the insecurity generated by it have been another cause of pastoralists’ migration. Table 3 presents probit model results of the deter- minants of migration amongst pastoral communities in Baringo District, estimated using the probability of a herder migrating. The log likelihood ratio statistics [LR chi 2 (12) = 61.22] indicate that the model fits the data significantly at 1% level. The majority of the households surveyed were headed by men (Figure 1). The find- ings of the present study indicate that genderofthehouseholdheadisasignificant determinant of migration (P < 0.1) (Table 3). Households that are headed by males are more likely to migrate than those headed by females. In addition to gender, the age of the household head also has a negative and significant (P < 0.05) effect on migration. The level of education of the household head represented by the number of years in school is not a significant determinant of migration decision (Table 3). Similarly, the size of the household is also not a significant determinant of the decision to migrate. Table 2 Household characteristics of the pastoralists surveyed in Baringo District, Kenya Description Minimum Maximum Mean SD Household size 3 36 13.55 8.12 Age of household head (years) 18 72 44.04 13.11 Number of years in school 0 16 3.03 4.68 Number of cattle owned 0 702 65.56 105.33 Number of shoats a owned 0 3360 167.14 425.24 Ratio of shoats a to cattle 0 9.00 2.58 2.07 Non-livestock income (KES) 0 840,000 24,167.27 88,937.63 Value of livestock lost to rustling (KES) 0 750,000 90,910.00 127,766.52 a Sheep and goats. KES, Kenya Shillings (80 KES = US $1); SD, standard deviation. Table 3 Determinants of the decision to migrate amongst pastoral communities in Baringo District, Kenya Variable Coefficient S.E. P >z Effects Gender of household head 0.188 a 0.690 0.085 0.008 Age of household head -0.040 b 0.019 0.035 -0.001 Education level of household head 0.015 0.047 0.741 0.001 Size of the household 0.011 0.039 0.780 0.001 Number of cattle owned 0.021 b 0.010 0.032 0.001 Ratio of sheep and goats to cattle -0.095 0.130 0.467 -0.003 Log non-livestock income -0.222 b 0.112 0.048 -0.008 Intensity of cattle rustling 2.207 c 0.659 0.001 0.243 Type of land ownership 0.073 0.675 0.914 0.002 Drought and/or diseases 1.377 b 0.677 0.042 0.164 Livestock lost to rustlers 0.647 a 0.493 0.089 0.030 Herder’s perception on livestock migration 2.415 c 0.654 0.001 0.255 Constant term -3.240 1.333 0.015 Number of observations 110 LR chi 2 (l2) 61.22 Probability > chi 2 0.000 Log likelihood -25.790 Pseudo R 2 0.543 a Significant 10%; b significant at 5%; c significant at 1%. S.E., standard error. Kaimba et al. Pastoralism: Research, Policy and Practice 2011, 1:18 http://www.pastoralismjournal.com/content/1/1/18 Page 7 of 16 The number of cattle owned by a pastoralist increases the probability of a herder migrating (Table 3). This is shown by the positiv e and significant (P <0.05)impactof number of cattle on migration. The estimates indicate that increasing the number of cattle owned by pastoralists by 10% would increase the probability to migrate by approximately 0.01%. On the other hand, the ratio of sheep and goats to the number of cattle has a negative coefficient. Non-livestock income has a negative and significant (P < 0.05) influenc e on migration. Increasing non-livestock income by 10% would decrease the probability to migrate by 0.08%. This study captured the in fluence of cattle rustling on migration by use of two dummy variables that have showed different reactions to cattle rustling occurrences. The intensity of cattle rustling influences herders’ decision to migrate po sitively and significantly (P < 0.01). This implies that severe, the herders are likely to migra te with their herd to safer areas to avoid loss from cattle rustlers. Likewise, t he variable on whether a herder has lost livestock to cattle rustlers also has a positive and significant (P < 0.1) influence on migration. Herders that have lost livestock to cattle rustlers in the past are more likely to migrate due to cattle rustling or the threat of it than those who have not lost livestock before. The migration decision caused by factors related to cattle rustling is taken as a form of insurance against the vice. The majority of the pastoralists in Baringo District own land on a communal basis. However, results indicate that this type of land ownership is not an important determi- nant of the decision to migrate (Figure 1 and Table 3). On the other hand, loss of live- stock to drought and/or diseases or any other biophysical factor has a positive and significant (P < 0.0 5) effect on migration. Herders that have lost livestock to drought and/or diseases before are more likely to migrate in search of water and pasture or flee from diseases and insecurity than their counterparts who have not. Besides, herders ’  0 20 40 60 80 100 Male Female Educated Not educated Severe Moderate Self Hired Communal Private Gender of household head Education level of household members Cattle rustling intensity Labour used by household Household land ownership % respondents Figure 1 Percent respondents to household characteristics amongst the pastoralist sampled in Baringo District, Kenya. Kaimba et al. Pastoralism: Research, Policy and Practice 2011, 1:18 http://www.pastoralismjournal.com/content/1/1/18 Page 8 of 16 perception of livestock migration influences migration decision positively and signifi- cantly (P < 0.01). Herders that perceive migration positivel y are more likely to migrate for whichever reason than those who perceive it negatively. Determinants of pastoral herd size The impacts of cattle rustling, migration and other socioeconomic factors were tested through their influence on herd sizes. The results of the single-equat ion estimation of herd size is presented in Table 4 while the three-stage least squares (3SLS estimation of herd size is presented in Table 5. The results for both single-equation estimation and 3SLS methods are compar ed very closely, indicating that there is no simultaneity. The Chow tests (F statistics) for all the specification co nfirm the goo dness of fit of the model and confirm the stability of the coefficients to changes in specification. Results on estimation of herd size indicate that gender of the household head has a posi- tive and significant (P < 0.1) influence on herd size, which means that households that are headed by males are more likely to keep larger herds than those headed by females. More- over, the age of the household head positively and significantly (P < 0.1) influences the herd size, such that elderly household head are more likely to keep bigger herds than their younger counterparts. The level of education has a negative and significant (P < 0.1) influ- ence on herd size, suggesting that herders with higher education levels are more likely to keep fewer numbers of livestock than those with lower education levels. Similarly, the size of the household has a significant (P < 0.01) but positive impact on herd size. This implies that large households own larger herd sizes than small households. Non-livestock income exerts a strong negative and significant (P<0.01) impact on herd size. There is an inverse relationship such that when non-livestock income Table 4 Single-equation regression analysis for herd size determinants amongst pastoral communities in Baringo District, Kenya Variables Coefficient S.E. zP>z Gender of household head 0.214 a 0.124 1.72 0.089 Age of household head 0.007 a 0.004 1.87 0.065 Education level of household head -0.017 a 0.008 -1.97 0.051 Size of the household 0.024 c 0.006 4.11 0.002 Log non-livestock income -0.070 c 0.019 -3.66 0.001 Intensity of cattle rustling -0.001 b 0.093 -0.01 0.039 Type of land ownership 0.029 0.162 0.18 0.858 Drought and/or diseases -0.184 a 0.167 -1.10 0.075 Livestock lost to rustlers -0.001 0.084 -0.00 0.997 Livestock inherited 0.436 b 0.172 2.54 0.013 Dowry received 0.086 0.094 0.92 0.362 Livestock bought 0.116 0.080 1.45 0.149 Predicted probability of migrating 0.436 c 0.161 2.70 0.008 Constant term 0.316 0.258 1.23 0.223 Number of observations 110 F(13,96) 7.18 Probability >F 0.000 R-squared 0.493 Adjusted R-squared 0.424 RMSE 0.383 a Significant at 10%; b significant at 5%; c significant at 1%. S.E., standard error. RMSE, root mean square error. Kaimba et al. Pastoralism: Research, Policy and Practice 2011, 1:18 http://www.pastoralismjournal.com/content/1/1/18 Page 9 of 16 increases by 10%, herd size is likely to decreases by 0.7%. Also, cattle rustling intensity has a negative and significant (P<0.05) influence on herd size, indicating that when- ever cattle rustling intensity moves towards severity, the pastoralists are more likely to reduce their herd size. Similarly, though not significant, the coefficient for livestock lost to cattle rustlers is negative in determination of the herd size. The predicted prob- ability of migrating has a significant (P<0.01) positive influence on he rd size, suggest- ing that herders who migrate are likely to have larger numbers of livestock than those who do not migrate. Drought and diseases influences herd size negatively. This is shown by the significant (P<0.1 ) influence the coefficient of drought and diseases has on herd size (Table 4), implying that tho se that have lost livestock to drought and diseases previously are more likely to own smaller herds than those not affected. In contrast, livestock inheri- tance showed a very significant (P<0.01) and p ositive influence on herd size. Herders who have inherited livestock are likely to have larger herds than those who have not. Furthermore, results indicate that the majority of households have at one time or another inherited livestock from their relatives. On the contrary, both dowries received and livestock bought did not significantly influence pastoralists’ herd size. Discussion Determinants of pastoral migration decisions The observation in the present study where male-headed households are more likely to migrate is in agreement with the traditional/cultural norms of most African pastoralists that allocate the responsibility to decide where to locate the household to the husband. These results are also consistent with the traditional model of household decision mak- ing reported by (Doss and McPeak (2005)), where husbands make decisions about herd Table 5 Three-stage least squares regression analysis for herd size determinants amongst pastoral communities in Baringo District, Kenya Variables Coefficient S.E. zP>z Gender of household head 0.212 a 0.116 1.83 0.068 Age of household head 0.006 b 0.003 1.98 0.048 Education level of household head -0.018 b 0.008 -2.11 0.035 Size of the household 0.024 c 0.005 4.41 0.001 Log non-livestock income -0.070 c 0.018 -3.90 0.001 Intensity of cattle rustling -0.003 b 0.087 -0.04 0.471 Type of land ownership 0.031 0.151 0.20 0.840 Drought and/or diseases -0.186 a 0.156 -1.19 0.083 Livestock lost to rustlers -0.001 0.079 -0.01 0.989 Livestock inherited 0.438 c 0.160 2.74 0.006 Dowry received 0.089 0.088 1.01 0.311 Livestock bought 0.115 0.074 1.55 0.122 Predicted probability of migrating 0.427 c 0.151 2.84 0.005 Constant term 0.316 0.241 1.31 0.190 Observations 110 Probability >F 0.000 RMSE 0.358 R-squared 0.49 Chi 2 106.78 a Significant at 10%; b significant at 5%; c significant at 1%. S.E., standard error. RMSE, root mean square error. Kaimba et al. Pastoralism: Research, Policy and Practice 2011, 1:18 http://www.pastoralismjournal.com/content/1/1/18 Page 10 of 16 [...]... Through migration, herders may also be able to avoid insecurities brought about by cattle rustling Conclusions and implications This study intended to elucidate the effects of cattle rustling and other household characteristics on migration decisions and herd size amongst the pastoralists in Baringo District in Kenya Gender and age of the household head are important determinants of the decision to migrate... cycle of movement is determined not only by availability of pasture and water but also by the varying seasonal patterns of disease (Raikes 1981) The present observation where the type of land ownership is not a significant determinant of migration decisions may be attributed to most pastoralists in Baringo not having individual ownership of land but rather depending on communal lands In such communal lands,... for migration in search of range resources In addition, small ruminants are not as fast as cattle in terms of mobility and take more time during migration, which is a limitation in case the herders are being pursued or are intending to move for long distances As a result, it is easier and faster to migrate with cattle than with sheep and goats Pastoralists engaged in non-livestock income-generating... chances of migrating, implying they are less likely to migrate than their younger counterparts Results from the present study indicate that the household head’s level of education and the size of the household are not a significant determinant of migration decision It is probable that the causes of migration may be affecting all households, regardless of the level of education of the heads and household sizes... rearing are likely to keep smaller herds of animals This is perhaps an indication that herders may not invest their non-livestock income into increasing their herd size It could also imply that livestock might be sold in order to invest in other non-livestock activities As explained previously, pastoralists may reduce their herd for various reasons, including dowry payment, fear of losing animals to insecurity... through raiding Thus, the insecurity associated with raiding leads to migration and the escape may involve long or short distances, depending on the information available about the level of insecurity and the availability of resources ((Young et al 2005)) The route followed and the length of stay will depend on the intensity of the rustling It is known that cattle rustling leads to loss of livestock,... whereas household heads with higher level of education are more likely to keep smaller herd sizes Also, households with bigger family sizes and those that have inherited livestock are more likely to own larger herds of livestock However, generation of income outside livestock rearing by the pastoralists leads to the keeping of smaller herds of animals The intensity and frequency of cattle rustling inversely... through migration from one locality to another The herder’s perception of livestock migration is quite important in determining the decision to migrate Those who perceive migration positively see it as a better means of survival for the livestock (Kabubo-Mariara 2003) The pastoralists in Baringo District, particularly the Pokot community, usually migrate in search of pasture and water during the dry season... migrate and herd size Households headed by younger males are more likely to make migratory decisions Also, the ownership of large number of cattle and the occurrence of droughts and diseases influences pastoralists decision to migrate However, the engagement in non-livestock income-generating activities reduces the possibility of migration Male-headed households are more likely to keep larger herds of. .. activities of cattle rustling influences decision making by pastoralists, a view supported by (Hendrickson et al (1996)) and (Mkutu (2006)) Though not significantly influencing herd size, the coefficient for livestock lost to cattle rustlers suggest that it has far-reaching repercussions on herders’ decision-making process For example, if a herder decides to migrate in fear of attack or as a result of an . Access Effects of cattle rustling and household characteristics on migration decisions and herd size amongst pastoralists in Baringo District, Kenya George K Kaimba 1 , Bernard K Njehia 2,4 and. intended to elucidate the effects of cattle rustling and other household characteristics on migration decisions and herd size amongst the pastoralists in Bar- ingo District in Kenya. Gender and. Kaimba et al.: Effects of cattle rustling and household characteristics on migration decisions and herd size amongst pastoralists in Baringo District, Kenya. Pastoralism: Research, Policy and Practice